One Gram of Fish Oil a Day Slowed Biological Aging by Three to Four Months in a Randomized Trial. The Same Trial Missed All Six of the Endpoints It Was Built to Answer.
Most longevity interventions are expensive, unapproved, or supported by mouse data and enthusiasm. Omega-3 is none of those things. It costs less than a gym membership, sits on every pharmacy shelf, and in February 2025 it became the first supplement to slow a next-generation epigenetic clock inside a randomized, double-blind, placebo-controlled trial. That is a genuinely unusual sentence to be able to write. It is also a sentence that arrives attached to a trial which, five years earlier, had reported that the same supplement did nothing for blood pressure, fractures, physical performance, cognition or infection rates in the same people. This article is about how both of those things can be true, what a three to four month shift on a methylation clock is actually worth, and what a man in his forties or fifties should take from a result generated in a cohort with a mean age of 75.
Key Takeaways: What the DO-HEALTH Omega-3 Evidence Actually Shows
DO-HEALTH was a three-year, double-blind, placebo-controlled randomized trial in 2,157 community-dwelling adults aged 70 and older across seven centers in Switzerland, Germany, Austria, France and Portugal, testing vitamin D3 at 2,000 IU per day, marine omega-3 at 1 gram per day, and a 30-minute home strength program three times a week in a two-by-two-by-two factorial design. The primary results, published in JAMA in November 2020, found no benefit on any of the six primary endpoints: systolic blood pressure, diastolic blood pressure, non-vertebral fractures, the Short Physical Performance Battery, the Montreal Cognitive Assessment, or rate of infections. A post hoc analysis of DNA methylation in 777 Swiss participants, mean age 75.5 years and 59.7 percent women, was published in Nature Aging on 3 February 2025 by Bischoff-Ferrari and colleagues, and reported that omega-3 slowed PhenoAge by 0.16 standardized units, GrimAge2 by 0.32 units and DunedinPACE by 0.17 units, corresponding to roughly 2.9 to 3.8 months over three years. Vitamin D alone and the exercise program alone produced no measurable effect on any clock, and the additive benefit of combining all three interventions appeared on PhenoAge only, reaching 0.32 units. Earlier DO-HEALTH analyses reported secondary findings that did move, including a 61 percent lower risk of invasive cancer with all three treatments combined at an adjusted hazard ratio of 0.39, 39 percent lower odds of becoming prefrail, and omega-3 reductions of 13 percent in infections and 10 percent in total falls. VITAL, a trial ten times larger at 25,871 US adults using the same 1 gram omega-3 and 2,000 IU vitamin D doses, missed both of its primary endpoints of invasive cancer and major cardiovascular events. No trial has demonstrated that moving an epigenetic clock in adulthood produces a longer life, and no analysis in DO-HEALTH examined testosterone, fertility or any male-specific outcome.
What was the DO-HEALTH trial and what did it actually test?
DO-HEALTH was a three-year, double-blind, placebo-controlled randomized trial that enrolled 2,157 relatively healthy, community-dwelling adults aged 70 and older and tested three cheap interventions at once. Participants were recruited between December 2012 and November 2014 at seven centers across Switzerland, Germany, Austria, France and Portugal. The three interventions were vitamin D3 at 2,000 international units per day, marine omega-3 at 1 gram per day supplying EPA and DHA in roughly a one-to-two ratio, and a simple home strength exercise program of 30 minutes three times a week, with a flexibility program serving as the exercise control rather than nothing at all. The design detail that matters most is the two-by-two-by-two factorial structure. Every participant was randomized independently on all three interventions, which produces eight combinations and lets the trial ask not only whether each treatment works on its own but whether the effects stack. Very few longevity trials are built this way, and it is the reason DO-HEALTH can say anything at all about combinations. The cohort also deserves attention before any result is interpreted. These were not frail nursing home residents. They were active, independent adults over 70 with no major disability at entry, and roughly four in five had adequate vitamin D status already. A population with little room to deteriorate over three years is a population in which a supplement has little room to demonstrate benefit, which becomes relevant the moment the primary results are read.
Did the trial's primary endpoints show any benefit?
No. DO-HEALTH reported no statistically significant benefit from vitamin D, omega-3 or the exercise program on any of its six primary endpoints. The primary results appeared in JAMA in November 2020, volume 324, issue 18, and the six endpoints were systolic blood pressure, diastolic blood pressure, incident non-vertebral fractures, the Short Physical Performance Battery as a measure of lower limb function, the Montreal Cognitive Assessment as a measure of cognition, and the rate of infections. All six came back null. This is the single most important fact about DO-HEALTH and it is almost never mentioned alongside the biological aging headline. There are two honest readings of a null result in a trial like this, and both are worth holding at once. The first is that the interventions do not work for these outcomes at these doses over three years, which is the plain reading and the one a statistician would default to. The second is that a healthy, active, vitamin D replete cohort tested for three years is close to the hardest possible setting in which to show benefit, because the participants were not deteriorating fast enough for a treatment to slow anything measurably. Neither reading rescues the interventions. What both readings do is set the bar for how much weight the secondary and post hoc analyses can carry, and that bar is lower than a headline about biological age reversal implies.
Six Primary Endpoints, Zero Hits
What DO-HEALTH reported on outcomes a person can actually feel, expressed as percent reduction versus control. Dark grey marks the six endpoints the trial was designed and powered to answer, published in JAMA in November 2020: blood pressure, fractures, physical performance, cognition and infection rate all came back null. Amber marks the secondary and follow-up analyses that did move, including a 61 percent lower risk of invasive cancer with all three treatments combined at an adjusted hazard ratio of 0.39, 39 percent lower odds of becoming prefrail, and omega-3 reductions of 13 percent in infections and 10 percent in total falls. Secondary findings from a trial that missed its primaries are hypotheses, not conclusions.
How much did one gram of omega-3 a day slow biological aging?
Omega-3 slowed three of six DNA methylation clocks by 0.16 to 0.32 standardized units over three years, which the authors translated as roughly 2.9 to 3.8 months of averted biological aging. The analysis was published in Nature Aging on 3 February 2025 by Bischoff-Ferrari and colleagues and covered 777 Swiss participants from the wider DO-HEALTH cohort, with a mean age of 75.5 years and 59.7 percent women, using blood drawn at baseline and again at year three. Six clocks were examined. PhenoAge slowed by 0.16 units, equivalent to about 2.9 months. GrimAge2 slowed by 0.32 units, about 3.8 months, and was the largest single effect in the study. DunedinPACE, which estimates the current rate of aging rather than an accumulated age, slowed by 0.17 units, corresponding to roughly a 1 percent reduction in pace. The original GrimAge did not move, and neither did the first-generation Horvath or Hannum clocks. That split is the most interesting part of the result and the part most often dropped. Second and third generation clocks such as PhenoAge, GrimAge2 and DunedinPACE were trained on mortality and health-span data rather than on chronological age alone, so they are more sensitive to the kind of inflammatory and metabolic signals omega-3 plausibly touches. Reading that as confirmation is one option. Reading it as a reminder that the answer depended on which of six instruments was used is the more cautious one, and the effect sizes themselves, between 0.16 and 0.32 standardized units, are small by any conventional benchmark.
Three Clocks Moved, Three Did Not
Standardized effect of 1 gram per day of marine omega-3 on six DNA methylation estimates of biological aging over three years, from the DO-HEALTH analysis published by Bischoff-Ferrari and colleagues in Nature Aging on 3 February 2025. Bars show the size of the slowing, so taller means more aging averted. GrimAge2 moved most at 0.32 standardized units, roughly 3.8 months. PhenoAge moved 0.16 units, roughly 2.9 months. DunedinPACE moved 0.17 units, about a 1 percent reduction in the pace of aging. The first-generation Horvath and Hannum clocks and the original GrimAge did not move at all, which is worth holding onto: this is a result about which clock you look at as much as it is a result about fish oil.
- Slowing versus placebo
Do vitamin D and exercise add anything on top of omega-3?
On one clock out of six, yes. On the other five, no. The DO-HEALTH methylation analysis found that combining omega-3 with vitamin D, with the exercise program, or with both, produced larger effects on PhenoAge than omega-3 alone, moving from 0.16 standardized units up into a range of 0.24 to 0.32 units. No additive effect appeared on GrimAge, GrimAge2 or DunedinPACE. Vitamin D alone and exercise alone produced nothing measurable on any of the six clocks. The finding that exercise did nothing needs a caveat that the headlines omit. The exercise intervention in DO-HEALTH was a 30-minute home strength routine performed three times a week, and the comparison group performed a flexibility program rather than sitting still. That is a low training dose measured against an active comparator in adults over 70, which is a very different proposition from testing progressive resistance training against genuine inactivity. A null result under those conditions says something narrow about that specific program and almost nothing about strength training as an aging intervention, where the observational and mechanistic case remains substantial. The vitamin D null is more straightforward, because the cohort was largely replete at baseline and supplementing a nutrient somebody already has enough of is not expected to help. The practical implication of the additive finding is modest: it is one clock, in a post hoc analysis, and the difference between the single and triple combination amounts to a fraction of a standardized unit.
The Only Clock That Stacked
Reported effect on PhenoAge by treatment combination in the DO-HEALTH methylation analysis. Omega-3 on its own produced 0.16 standardized units of slowing. Adding vitamin D or the home exercise program pushed the combinations into a range running from 0.24 to 0.32 units, with all three together at the top of that range. Vitamin D alone and exercise alone produced nothing measurable on any clock in this cohort. The additive pattern appeared on PhenoAge only. GrimAge, GrimAge2 and DunedinPACE showed no additive effect across the interventions, so the stacking story rests on a single clock out of six.
Why did VITAL, a trial ten times larger, find nothing?
VITAL asked a different question and answered it with hard clinical endpoints, which is why its null result carries more weight for disease risk than DO-HEALTH's methylation result carries for aging. VITAL randomized 25,871 US adults, men aged 50 and older and women aged 55 and older including 5,106 African American participants, in a two-by-two factorial design using the same doses DO-HEALTH used: 2,000 IU of vitamin D3 and 1 gram of marine omega-3 per day. The results were published in the New England Journal of Medicine in 2018. Neither supplement reduced the primary endpoints of total invasive cancer or major cardiovascular events, defined as a composite of myocardial infarction, stroke and cardiovascular death. The two trials are not really in conflict, because they measured different things. VITAL counted diagnoses and deaths across roughly five years in a cohort large enough to detect small differences in those counts. DO-HEALTH measured methylation patterns in blood in 777 people and reported a shift in a statistical estimate. A molecular change can be real and still fail to produce fewer heart attacks, and VITAL is the trial that tested whether the heart attacks show up. When a large trial with clinical endpoints and a small analysis with molecular endpoints point in different directions, the sensible ranking puts the clinical endpoints first and treats the molecular result as a mechanism worth studying rather than a benefit already banked.
Is there a downside to one gram of omega-3 a day, especially for men?
The two risk signals worth knowing about are atrial fibrillation at higher doses and an observational prostate cancer association, and neither is settled. On atrial fibrillation, a 2021 meta-analysis in Circulation pooling long-term marine omega-3 cardiovascular outcome trials reported roughly a 25 percent higher risk of incident atrial fibrillation with supplementation. A larger 2025 update covering 34 randomized trials and 114,326 people found the signal was not uniform: it reached significance only in participants already at high cardiovascular risk who were taking more than 1,500 mg of EPA and DHA per day, where the pooled odds ratio was 1.48 with an absolute risk difference of 0.8 percent. In the other strata, including low dose regimens and people without established cardiovascular disease, the association did not reach significance. A 1 gram per day dose in a man without cardiac disease sits in the lower risk part of that picture, but a man with a history of atrial fibrillation or established heart disease is in the group where the signal appeared. The prostate question comes from Brasky and colleagues in the Journal of the National Cancer Institute in 2013, a case-cohort analysis inside the SELECT trial covering 834 prostate cancer cases, 156 of them high grade, against a 1,393 man subcohort. Men in the highest quartile of plasma phospholipid long-chain omega-3 had a hazard ratio of 1.43 for total prostate cancer, 1.44 for low-grade and 1.71 for high-grade disease, with the high-grade confidence interval running from 1.00 to 2.94. That analysis measured blood levels rather than supplement assignment, was heavily criticised on methodological grounds, and has not been reproduced by randomized supplementation trials including VITAL. It remains a reason for men to raise the subject with a physician rather than a reason to conclude anything. This article is educational and is not medical advice.
The Signals That Run the Other Way
Risk ratios above 1.0 associated with higher omega-3 exposure, with the dashed line marking no association. The prostate figures come from Brasky and colleagues in the Journal of the National Cancer Institute in 2013, comparing the highest to the lowest quartile of plasma phospholipid long-chain omega-3 in 834 prostate cancer cases from the SELECT trial: hazard ratio 1.43 for total prostate cancer, 1.44 for low-grade and 1.71 for high-grade disease, the last with a confidence interval running from 1.00 to 2.94. Those are blood levels in an observational case-cohort analysis, not a randomized supplement assignment. The atrial fibrillation figures come from randomized cardiovascular outcome trials, where the 2021 Circulation meta-analysis reported a 25 percent higher risk overall and a 2025 update of 34 trials in 114,326 people found the signal concentrated at doses above 1,500 mg per day in people already at high cardiovascular risk.
What does a slowed epigenetic clock mean, and what should a man over 40 take from this?
A slowed epigenetic clock is a change in a statistical estimate built to correlate with age and mortality across large populations, and no trial has yet shown that moving one in adulthood produces a longer or healthier life. That gap sits underneath every biological age headline, including this one. PhenoAge, GrimAge2 and DunedinPACE were trained to predict health outcomes, so a shift in them is a reasonable hypothesis about future risk. A hypothesis is not a result. The DO-HEALTH authors were candid about the limits: there is no gold standard for measuring biological aging, two blood draws three years apart carry substantial measurement error, and nobody knows whether the change persists after supplementation stops. For men specifically the gaps are wider than the trial's press coverage suggests. The methylation cohort had a mean age of 75.5 years and was 59.7 percent women, which makes it a poor guide to what a 45 year old man should expect. No DO-HEALTH analysis examined testosterone, sex hormone binding globulin, fertility or body composition, so anyone hoping the trial speaks to male hormonal health is asking a question it did not address. The practical position that follows is unusually simple for this field. Omega-3 at the trial dose of 1 gram per day of combined EPA and DHA costs roughly 50 to 150 dollars a year, has decades of safety data at that dose, and is obtainable from two or three servings of oily fish a week without a capsule. The reasonable reading is that this is a low cost, low risk habit with a plausible and now randomized mechanistic signal behind it, not a treatment that has been shown to extend anyone's life. Anyone with atrial fibrillation, established cardiovascular disease, a bleeding disorder, upcoming surgery or anticoagulant medication should discuss it with a physician before starting.
The Omega-3 Aging Record on One Axis
Every strand of the omega-3 biological aging case arranged as a single schematic map. Blue marks what the evidence supports, amber marks design limits that are real without being disqualifying, slate marks findings that count against the intervention, and dark grey marks claims the evidence does not carry. Omega-3 is the rare longevity candidate that is cheap, widely available, tested inside a real randomized placebo-controlled trial, and safe enough that the argument is about magnitude rather than harm. It is also a result from a post hoc analysis of 777 people with a mean age of 75.5 years, sitting inside a trial that missed all six of the endpoints it was built to answer.
Frequently Asked Questions
Have more questions? Contact us or explore our full peptide intelligence platform.
Related Articles
Does Telomere Length Determine Lifespan? Short Telomeres Track With Earlier Death. The Genetics Say Longer Ones Raise Cancer Risk.
Does telomere length determine lifespan? Short telomeres track with earlier death, yet genetic studies show longer ones raise the risk of several cancers.
Low-Dose Aspirin Added 8 Percent to Male Mouse Lifespan. In 19,114 Healthy Older Adults, It Added No Healthy Years and Deaths Ran Higher.
Does low-dose aspirin help longevity? It extended male mouse lifespan 8%, but the ASPREE trial in 19,114 healthy older adults found no gain and more bleeding.
Is NAD+ a Peptide? No. It Is a Coenzyme Built From Two Nucleotides, and the Peptide Label Comes From the Clinic, Not the Chemistry
Is NAD+ a peptide? No. NAD+ is a coenzyme built from two nucleotides. What NAD+ is used for, what IV NAD+ studies found, and how it differs from SS-31.